Re: [boost] Boost.Filesystem i18n mini-review

From: Beman Dawes (bdawes_at_[hidden])
Date: 2005-07-24 20:13:25

At 04:42 AM 7/20/2005, Jan Hermelink wrote:

 >This is not (yet) a review, but I use your filesystem library in several
 >unicode projects and am very interested in the wide char implementation.
 >I have tried wpath and wpath_directory_iterator with two projects and
 >run in several minor problems:
 >
 >- Building
 >I can't build the DLLs on my system (Windows XP SP2, VC 7.1):
 >  path.cpp(51) : error C2491: 'boost::filesystem::wpath_traits::imbue' :
 >  definition of dllimport function not allowed
 >
 >Static builds run fine.

This should be fixed soon. Thanks for the report.

 >- fstream.hpp:
 >  Declaration of function narrow_path(const std::wstring & file_ph, ...)
 >If several filesystem fstream objects are declared in different
 >translation units, the linker complains this symbol as multiple defined.
 >
 >Adding an inline specifier helps as workaround.
 >
 >- fstream.hpp:
 >  Implementation of function narrow_path(const std::wstring & file_ph,
 >...)
 >This function doesn't work at all on my system, it returns always the
 >empty string. Why is this 8.3 conversion needed,

It is a hack. To avoid having to modify the standard library (which doesn't 
support wide-character filenames), the 8.3 names are used instead. These 
appear to always be narrow even if the full name is wide.

 > can this work on NTFS reliable?

Apparently not. I wasn't aware you could turn them off.

 >Replacing the current implementation thru a simple narrow conversion
 >helps as workaround.

My understanding (which could be faulty) is that there is no general way to 
do such a wide to narrow conversion that always works. I'd love to be 
proven wrong.

 >- is_symlink()
 >Why isn't this function implemented on Windows? A NTFS junction is
 >symlink, isn't it?

So I've been told. I tried a couple of times to figure out how to use 
junctions, but my eyes glazed over long before understanding occurred.

 >After fixing the above problems the library seems to run fine.
 >
 >Thank you for your work!

And thank you for your comments!

--Beman